Use of Sentinel-2 Images for the Detection of Sandbars Along the Lower Vistula

Abstract

The aim of this paper was to develop and test detection methods of sandbars for a selected section of the Lower Vistula (Wisła) river with the use of Sentinel-2 Level 2A optical images. Material and methodsThe analyses were performed in the QGIS (version 2.18.4) and SNAP (version 7.0) software. Both the image reprogramming and pixel value analysis were conducted in the SNAP software. QGIS was used to perform satellite data, i.e. processing, surface analysis, and their visualisation. For multispectral images, water indices were used such as: NDWI (Normalized Difference Water Index); MNDWI (Modified Normalized Difference Water Index); AWEIsh (Automated Water Extraction Index shadow); AWEInsh (Automated Water Extraction Index no shadow); LSWI (Land Surface Water Index); MLSWI (Modified Land Surface Water Index); MSI (Moisture Stress Index); SWM (Sentinel Water Mask) to separate sandbars from a water. These index methods base on a threshold value. Not all tested indices provided satisfactory results. Therefore, the layers were generated for 5 water indices. Results and conclusionsThe analyses have shown that, for the selected section of the Lower Vistula, it is possible to detect sandbars in the river channel on the basis of Sentinel-2 satellite's data. A proper selection of remote sensing index, creation of binary classification and selection of processing algorithm are important when detecting river sand bars. Every water index used produced different results. The difference may be reduced by improving the threshold value algorithm.